SUBSCRIBE TO OUR NEWSLETTER
Stay informed! Sign up to get expert advice and insight delivered direct to your inbox
Implementing Automation
Orchestration platforms have better chances of success now that vendors are offering APIs to effectively manage their products. Routers, virtualization platforms, firewalls and storage arrays are delivering core features to enable external administration using XML-based APIs.
October 21, 2011
5 Min Read
WHITHER ORCHESTRATION?
Automation and Orchestration
Vendors are focusing on orchestration and automation of their infrastructure to extract pounds instead of pennies from their customers.
The Case Against Orchestration
Management platforms have failed to deliver automation and orchestration for the last 20 years at least. Why would today's software be any more likely to actually work?
In Automation and Orchestration, we spoke about what is automation, orchestration and what the vendors hope to sell. In The Case Against Orchestration, we considered why automation hasn’t been successful in the past and considered why it might be the right time for adoption. If you decide that’s it time to try again, let's consider what options, tips and plans you could make for a successful deployment.
Orchestration platforms have better chances of success now that vendors are offering APIs to effectively manage their products. Routers, virtualization platforms, firewalls and storage arrays are delivering core features to enable external administration using XML-based APIs.
Software vendors need a few years to mature their products before they will be truly useful across a wide range of products. An action for today is to ask equipment vendors about their support for scripting and automation capabilities with languages and XML APIs that you can use now.
Start Scripting Today
You may not be ready to purchase an automation or orchestration system today, but consider starting automation with some scripts. Scripting is an art that has faded in recent years as vendorsimproved product functionality and we got better software products. It’s not common today, but many IT people have Perl or TCL scripting capabilities, and more modern Windows and VMware administrators will have some PowerShell script knowledge. Set goals for your staff to develop a few scripts to automate simple tasks. Consider building a central Web console for these scripts to show status, completion and to provide useful feedback. This is the start of building your own automation system.
Realize that these scripts will need maintenance. At first, it will seem that this is "extra" work, but once automation has started to reduce the repetitive tasks, your team will have more time to work strategically and plan instead of doing repetitive tasks.
It's important to take time to check that the scripts take note of failure conditions and remember that validation is part of the development process. It takes extra time to check that the configuration was completed, valid and functional, but it will pay off in detecting when something has failed.
Having your own platform will help to develop business appreciation of automation and lead to easier acceptance of orchestration in the future.
Vendor Stacks
The trend for vendors to produce a "full solution stack" is all the rage this year. Consider the partnership of Cisco for data networking, storage networking and Wintel servers, and EMC for storage arrays that work closely with VMware. Or HP’s solution stack encompassed by its Matrix strategy that combines network, storage, Wintel servers and Microsoft HyperV--the list is quite long (and growing). Dell is building a single-source solution combining its acquisitions in Compellent/EqualLogic for storage, Force10 for network and its own server/blades, and all of this is happily independent of either Microsoft or VMware.
The value of these stacks is fully realized by the automation overlays that vendors are offering, where the best current examples are Cisco UCS Manager, HP Matrix products and IBM Tivoli products.
The single-vendor strategy creates opportunities to develop management software that finally works, and the narrow focus on selected products means the integration between the available APIs and the platform can be limited to a workable scope. One of the key failures of previous management tools was attempting to support a wide range of products to have a larger-sized target audience, at the price of deep capability and serviceability.
Consider the following ideas for improving your chances:
Achieve a "vendor stack" technology strategy using existing vendor partnerships for reduced integration risk
Choose an orchestration solution that is nominated for that vendor stack
Engage vendor resources for implementation
Do it early, test for success and ditch early. Expect it to fail. Current products are barely out of the version 1 beta and have a long way to go.
Lock the purchase decision with successful test and deployment; don’t pay for it up front--move the risk to a working deployment.
The People Tips
I think it is likely that the biggest challenge is changing the business practice and getting staff to automate their work. Some may realize it’s an opportunity to move up to more strategic work that has better outcomes, but most will resist the change. Good luck with that--it’s a classic management issue.Skills Management
You’ll need to develop new skills and daily tasks for your staff. This will need significant investment in lab and test systems where orchestration can be validated. But these test systems are also useful for skill and talent building. This requires a change in approach, but, most importantly, engineers will need to self-manage their lab time.Lab time requires well-organized individuals who can manage soft deliverables. Compare this with service delivery, which can be managed by measuring outputs to the business. It’s an old challenge for engineers, team leaders and management in IT that harks back to '90s technology, when measurement was not possible so we turned to overly simple management tools.
Dependency
Your organization will become dependent on orchestration tools once deployed. Like any dependency, acceptance is the first step. Infrastructure engineers and managers need to change their approach to move from reactive service delivery to proactive planning and project management. Accept this change and find ways to break the dependency on "quick-hit delivery" and move to "planned process."Cottage Industry or Factory
The push for orchestration has historic parallels in the transition from craftsmen in cottage industries in the 1800s to factories and mass production in the 1900s. Today, individuals build hand-crafted systems that are often unique in their own way. As IT scales up, a production line that is supported by significant capital investment in machines and tooling can produce a higher quality product at a lower price.That’s the promise of orchestration. It requires investment and vision to achieve, but the outcomes are lower production costs and improved productive output.
Read more about:
2011You May Also Like
More Insights
